🌀 nonlinear sciences

Fluidic hysterons and memory in flow networks

This paper demonstrates that passive hydrodynamic feedback between viscous flow and elastic fiber deformation can create fluidic hysterons, enabling the construction of memory networks that transition from non-interacting return point memory to avalanche-like switching through geometric control.

Original authors: Abhineet Singh Rajput, Amir A. Pahlavan

Published 2026-07-17

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Abhineet Singh Rajput, Amir A. Pahlavan

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Secret Life of Flowing Water

Imagine a world where water doesn't just flow; it remembers. In the hidden corners of physics, scientists study "memory" not as a human trait, but as a mechanical one. Think of a rubber band: if you stretch it and let go, it snaps back. But some materials are trickier. They have a "bistable" nature, meaning they can sit comfortably in one of two distinct shapes, like a light switch that is either firmly "on" or firmly "off." When you push these materials, they don't change smoothly; they suddenly "snap" from one state to another. This behavior is called hysteresis. It's the reason a magnet stays magnetized after you remove it, or why a crumpled piece of paper doesn't unfold exactly the same way twice.

Usually, we see this in solid things—buckling metal, snapping plastic, or shifting grains of sand. But what about liquids? Water is famous for being reversible; if you stir it and then stir it back, it returns to its original state. However, scientists have long wondered: Can a network of flowing fluids, like the tiny blood vessels in your brain or the tubes in a slime mold, develop a memory of its own? Can the flow itself, interacting with flexible walls, create a system that "remembers" how it was pushed in the past? This is the question that sits at the heart of a new study from Yale University, exploring how water and wobbly fibers can team up to build a memory machine without any electronics or solid switches.

The Paper: When Water Snaps Like a Switch

In this study, researchers Abhineet Singh Rajput and Amir A. Pahlavan discovered a way to make a single strand of elastic fiber act like a memory switch, but entirely through the power of flowing water. They didn't use a pre-made "snap-through" spring or a buckling beam. Instead, they built a tiny microfluidic channel (a microscopic pipe) and anchored a soft, elastic fiber inside it at a slight angle.

Here is the magic trick: As they slowly increased the flow of water through the channel, the fiber didn't just bend gently. At a certain point, it suddenly snapped into a new position. Why? It's a game of feedback. As the water pushes the fiber, the fiber bends and narrows the gap next to it. This narrowing makes it harder for water to squeeze through that gap, which changes the pressure and pushes the fiber even harder, causing it to snap shut. This is a "fluidic hysteron"—a memory element born purely from the dance between flowing liquid and bending solid.

The researchers found that whether the fiber snaps or just bends depends on its geometry. If the fiber is short compared to the width of the channel, it bends smoothly. But if the fiber is long enough (specifically, when the ratio of its length to the channel width, $\lambda$ , is around 1), it undergoes a sudden, discontinuous jump. This jump creates a "hysteresis loop": the flow rate needed to snap the fiber shut is different from the flow rate needed to snap it open again. The system "remembers" if it was recently snapped or not.

The "Load Line": A Bypass for the Brain

To turn this single fiber into a tunable device, the team added a second, parallel channel—a "bypass"—next to the fiber channel. They could change the width of this bypass, controlled by a ratio called $\beta$ . Think of the bypass as a detour road.

When the bypass is narrow (or non-existent), the fiber does its thing: it snaps, creating a memory. But as the researchers widened the bypass, they found they could "tune" the memory. A wider bypass acts like a safety valve, stealing some of the water flow away from the fiber. This reduces the pressure on the fiber, making it harder to snap. If the bypass is wide enough, the fiber never snaps at all; the memory disappears. This allows the scientists to smoothly transition the fiber from a "snappy" switch to a "smooth" bend just by changing the width of a neighboring channel. It's like having a dimmer switch for memory itself.

The Avalanche: When One Fiber Drags the Others

The real fun begins when they put ten of these fibers in a row. In a simple setup with no bypass, each fiber acts independently. If you increase the flow, they snap one by one in a predictable order, like a row of dominoes falling. This is a "non-interacting" system, and it follows a strict rule called "Return-Point Memory" (RPM). This means if you push the system halfway, stop, and then pull it back to where you started, it returns to the exact state it was in before. It's a perfect, obedient memory.

However, when they opened the bypass (increasing $\beta$ ), the fibers started talking to each other through the water. Because they share the same pressure, when one fiber snaps and blocks its path, the water gets pushed into the bypass, changing the flow conditions for the other fibers. This creates a "long-range" interaction.

Suddenly, the orderly domino effect breaks. Instead of snapping one by one, the fibers start triggering avalanches. One fiber snaps, which changes the flow so much that it forces a second fiber to snap immediately after, or even causes a previously snapped fiber to "un-snap." The system jumps between states in chaotic, diagonal leaps rather than smooth steps.

Most importantly, this interaction destroys the Return-Point Memory. If you push the system halfway and then pull it back, it doesn't return to the exact same state. It gets "stuck" in a new configuration. The system has learned a new, more complex way of remembering—one that is messy, interactive, and full of surprises.

What This Means

The paper proves that you don't need complex electronics or solid-state switches to create memory. You can build it entirely out of passive, flowing fluids and flexible fibers. The "memory" emerges from the feedback loop: the flow bends the fiber, the bent fiber changes the flow, and the new flow pushes the fiber further.

The researchers showed that by simply adjusting the geometry of the channels (the width of the bypass), they could control whether the system acts like a simple, obedient switch or a complex, interacting network prone to avalanches. This suggests that in nature, perhaps in the tiny blood vessels of a brain or the tubes of a slime mold, similar fluid-structure feedback loops could be the reason these living networks "remember" their history and adapt their shapes over time. It turns out that even without a brain, water and wobbly walls can tell a story of where they've been.

Problem Statement
Memory in driven matter is often described by "hysterons"—bistable elements with distinct switching thresholds that generate hysteresis, avalanches, and return-point memory (RPM). While experimental realizations of hysteron networks have been dominated by solid-state mechanical systems (relying on buckling, snap-through, or geometric incompatibility), a passive fluidic counterpart has remained elusive. In microfluidic networks, viscous flows typically obey linear Stokes equations and are kinematically reversible, meaning history dependence usually requires active control, viscoelasticity, or external actuation. The central problem addressed is whether bistability and memory can emerge intrinsically in a passive fluidic system through flow–structure feedback, without requiring structurally bistable elements or external intervention.

Methodology
The authors realized fluidic hysterons using a microfluidic setup containing elastic fibers anchored within rigid channels.

Single Element: A single elastic fiber (length $L$ , width $b$ , height $h$ ) is anchored at an upstream tilt angle $\theta_0$ in a channel of width $W$ . As the volumetric flow rate $Q$ is ramped quasistatically, viscous loading deforms the fiber. This deformation narrows the lateral gap, increasing hydraulic resistance, which in turn redistributes flow and modifies the hydrodynamic loading. This nonlinear elastohydrodynamic feedback loop is modeled using a reduced elastohydrodynamic theory coupling lubrication-based hydraulic resistances with an inextensible Euler–Bernoulli beam description.
Geometric Control: To tune interactions, a parallel bypass channel of width $W_2$ is added alongside the fiber channel ( $W_1$ ). The ratio $\beta = W_2/W_1$ acts as a geometric load line, controlling the fraction of flow diverted away from the fiber channel.
Arrays: Experiments were conducted on arrays of $N=10$ fibers connected in series with a shared bypass. The system was subjected to standard write–read protocols involving nested turning points to test for return-point memory.
Fabrication: Devices were fabricated using soft lithography (PDMS) with in situ photopolymerization of PEGDA hydrogel fibers. Flow was controlled via syringe pumps to ensure negligible inertia ( $Re \ll 1$ ), and fiber deformation was tracked via brightfield microscopy.

Key Contributions and Results

Fluidic Bistability via Cusp Catastrophe: The study demonstrates that a single elastic fiber, which is not intrinsically bistable, becomes bistable through flow–structure feedback. Long fibers ( $\lambda = L/W \approx 1$ ) exhibit a discontinuous "snap" transition, whereas short fibers deform continuously. This transition is governed by a cusp catastrophe in the geometric control parameters ( $\lambda, \theta_0$ ). The onset of snapping follows a square-root scaling law ( $\Delta\theta \sim (\lambda - \lambda_c)^{1/2}$ ), and the hysteresis window is determined by the fold lines emerging from the cusp point.
Geometric Load-Line Control: The parallel bypass channel acts as a tunable load line. Increasing the bypass width ratio $\beta$ shifts the snapping threshold to higher flow rates, reduces the jump amplitude, and can completely suppress bistability. This allows the system to traverse the negative-differential branch of the constitutive curve (which is inaccessible in a single-channel flow-controlled setup), effectively converting the fiber into a passive flow router that switches between open and closed states.
Interacting Hysterons and Avalanches: In arrays, the bypass channel mediates long-range hydraulic interactions.
- In the non-interacting limit ( $\beta = 0$ ), fibers behave as independent hysterons, exhibiting standard Preisach behavior with return-point memory (RPM).
- In the interacting regime ( $\beta > 0$ ), the state of one fiber alters the effective resistance and flow seen by others. This leads to "antiferromagnetic-like" interactions where the snapping of one fiber disfavors the snapping of others.
- This coupling produces avalanche-like transitions (simultaneous switching of multiple fibers) and diagonal jumps in the phase space, violating the sequential switching of the non-interacting limit.
Violation of Return-Point Memory: As the interaction strength increases, the system transitions from a non-interacting Preisach regime to an interacting regime where RPM is violated. Minor loops extend beyond the major loop, and the system fails to recover previously visited intermediate states upon reversal of the drive. The number of avalanches per cycle increases with $\beta$ .

Significance
The paper establishes a passive hydrodynamic route to hysteron networks where memory emerges from flow–structure feedback and global hydraulic constraints rather than solid-state multistability or external control.

Theoretical Framework: It provides a minimal physical realization of the hysteron paradigm in fluids, demonstrating that bistability can arise from the coupling of viscous flow and elastic deformation, organized by a cusp catastrophe.
Interaction Mechanism: Unlike solid-state systems where coupling is mediated by short-ranged elastic stresses, this system features all-to-all coupling mediated by Kirchhoff-like hydraulic constraints. The interaction strength is tunable via a single geometric parameter ( $\beta$ ).
Biological Analogue: The results offer a minimal physical analogue for living flow networks (e.g., fungal mycelia, microvasculature), suggesting that history-dependent collective responses and memory can emerge from passive ingredients alone, without requiring active remodeling or biochemical signaling.
Design Principle: The work suggests that by arranging and coupling fluidic hysterons, one can encode mechanical information processing, such as counting and reproducible switching sequences, purely through geometric design and flow constraints.

The Secret Life of Flowing Water

The Paper: When Water Snaps Like a Switch

The "Load Line": A Bypass for the Brain

The Avalanche: When One Fiber Drags the Others

What This Means

More like this